Where AI comes from can be extremely complex. The all-encompassing term AI refers to a system of technologies that generally has different sovereignties. There’s the data that trains the model, the model itself, the compute and tooling used to run the model, and even the customer base — all this is likely to be located in different places.
The UK is very much a leader in AI, but sovereignty is a challenge. We have always been at the forefront of computational science, but that’s not to say we have a handle on the supply chain. Large tech companies have vast control over many elements, and they are not based here.
Data might be used to train models on machines within the UK, but those machines might be owned by a foreign entity that has invested in the UK. The data itself might come from anywhere and have complicated ownership.
Why does this matter? Because understanding sovereignty is key to having control over the value chain of your solution.
One element of this is financial. The government has announced huge investment into AI in the UK. That’s nothing compared to the huge amount of investment in AI infrastructure announced by the world’s leading tech companies. We need to keep hold of this spend and have it further boost our economy, rather than let money slip down the supply chain elsewhere.
For example, in the UK we’ve always been very good at developing novel hardware, such as Raspberry Pi or Arm’s semiconductor chips. There are a lot of people in the UK currently working on novel architectures to accelerate AI computation beyond what we can do with today’s GPUs.
But where will these devices be manufactured? The UK is not investing enough in semiconductor manufacturing. This puts vulnerability into our technologies.
And what will happen once it is manufactured? Without control over the supply chain, we also don’t have control over the value chain.
Generally, when you sell the device by itself, it’s not really useful. You need a series of system integrators to take that chip and deliver it to a customer in a way that is valuable, to host it, power it and so on. On top of that, you need service providers providing the software stack.
There’s a big difference between a very valuable technology and success in the market, and a lot of pieces that need to go together. As well as investment, one of the key questions is security. If your data is being processed on a machine owned by a foreign organisation, we don’t necessarily have full control of that data or compute.
It’s not about making sure that everything is British; it’s about making sure that we don’t have vulnerabilities. Sovereignty is about having choices. To achieve that, we need diversity in the value and supply chains to make sure no single country or organisation has a monopoly. New players with novel business models and technologies are essential to compete and disrupt a market, and we need to give AI organisations the ability to choose them.
CommonAI Compute is delivering a GPU marketplace so that AI organisations can be instantly multicloud. Rather than being tied to using one of the large, established cloud vendors they can easily access a range of providers.
In turn, this opens the door for smaller cloud providers — located anywhere in the world. If AI organisations can access these providers, they can be in control of their own supply chain again and understand their own sovereignty.
It’s unrealistic to believe we can create an AI sector where every element is based in the UK, but what is key is having control over our spend. The huge demand from AI for compute is spurring on new providers and I’m excited to see what innovations emerge.